|Home | About | Journals | Submit | Contact Us | Français|
Multiple models have been proposed to interpret the retention of duplicated genes. In this study, we attempted to compare whether the duplicates arising from tandem duplications and retropositions are retained by the same mechanisms in human and mouse genomes.
Both sequence and expression similarity analyses revealed that tandem duplicates tend to be more conserved, whereas retrogenes tend to be more divergent. The duplicability of tandem duplicates is also higher than that of retrogenes. However, positive selection seems to play significant roles in the retention of both types of duplicates.
We propose that dosage effect is more prevalent in the retention of tandem duplicates, while 'escape from adaptive conflict' (EAC) effect is more prevalent in the retention of retrogenes.
Gene duplication is one of the most important sources of genomic novelty and complexity . There are three main molecular mechanisms leading to new duplicates [2,3]: 1) unequal crossing-over during homologous recombination, 2) duplicative transposition at the DNA level and retroposition mediated by mRNA, and 3) polyploidization. While polyploidization is characterized by bursts of large-scale genome duplication, the former two processes are often small-scale and proceed continuously . Recently, the investigations of full genome sequences have revealed that both large- and small-scale duplications play significant roles in the evolution of various organisms . Although the molecular basis of gene duplication has been well understood, how the newly created duplicates are fixed in the population is still quite controversial . Several evolutionary models for this issue have been proposed, and according to the current perspective , they can be distinguished from two independent dimensions: 1) the extent of functional divergence for the new duplicates, and 2) whether positive (adaptive) selection is involved in the process. The outcomes of functional divergence are usually classified as gene conservation, subfunctionalization and neofunctionalization [2,3], though the definitions for the latter two are often ambiguous. Theoretically, the duplicates can undergo adaptive evolution or neutral genetic drift to achieve each outcome.
Statistical analyses on empirical data have suggested that none of the mechanisms alone can interpret the maintenance of all duplicates . However, we suspect that these retention mechanisms may not contribute equally for duplicates stemming from different molecular bases. In fact, by examining the substitution rate between duplicated pairs, Jun et al.  have found that retrotransposed and interspersed segmental duplicates diverge more quickly than tandem duplicates. To further compare the underlying retention mechanisms, we attempted to investigate the tandem duplicates arising from unequal crossing-over and retrogenes arising from retroposition in human and mouse genomes. We chose both types of duplicates because: 1) tandem duplicates and retrogenes are easier to screen, and 2) after ancient large-scale genome duplications at the origin of vertebrates, most duplicates have been created via small-scale events in mammalian genomes . In addition, we made the assumption that the duplication rate for each type is constant per year rather than per generation in mammalian genomes. This seems a reasonable assumption because the duplication rate has often been presented with respect to absolute time scale in previous studies [9,10].
All paralogs (protein-coding genes with pseudogenes excluded) and relevant annotations (identity scores, locations and exons) were retrieved from Ensembl database (release 50) via BioMart , which amounted to 80,683 and 159,047 pairs of duplicates in human and mouse genomes, respectively. The original dataset had a lot of redundant pairs in multi-member gene families. For example, in a n-member family, there would be n(n-1)/2 paralogous pairs listed, although at most only n-1 duplication events were needed to create the family. In this case, we only chose the n-1 pairs that contained all the members and had the highest total identity score. Altogether, 9,425 and 11,224 non-redundant pairs were preserved for the two genomes. Next, we applied CHSMiner  to detect and remove paralogous segments arising from large-scale duplications. The segments should contain at least two pairs of duplicates, and the gap size between two neighbouring duplicates in either segment should be less than 30 genes . The duplicated pairs located in those segments with FDR < 0.05 were filtered. After this step, we obtained 6,552 and 8,308 pairs for further screening in human and mouse genomes, respectively.
Although tandem duplicates should be adjacent to each other on one chromosome, the extensive gene inversions may insert irrelevant genes into the tandem arrays. We followed the stringent definition adopted by previous studies [14,15] to screen the tandem duplicates, which restricted the inserted spacers to no more than one gene. This resulted in 1,210 and 1,802 paralogous pairs in human and mouse genomes, respectively [see additional file 1 and 2]. We implemented a method similar to those of Emerson  and Pan  to screen retrogenes. First, the pairs with a multi-exon member and an intronless member were considered as putative parental-retrogene pairs, but the pairs with both members intronless were ignored as they were not clearly created via retropositions. Next, for the putative pairs with both members located on the same chromosome, we discarded those with the intervening spacers containing less than 10 genes, since they were confused with tandem duplicates. Finally, we preserved 410 and 680 pairs resulting from retropositions in the two genomes, respectively [see additional file 3 and 4].
The similarity of protein sequences between two duplicates, as measured by their average amino acid identity, can be retrieved directly from BioMart. The dN and dS of their coding sequences were downloaded from the EPGD database http://epgd.biosino.org/EPGD/ [see additional file 1, 2, 3, and 4]. To avoid the influence of saturation effect , only the pairs with dS < 1 were considered in the dN/dS analysis.
The tissue-specific expression profiles and the annotation of the probesets were downloaded from the GNF gene expression database http://wombat.gnf.org. We chose the datasets HUMAN U133A/GNF1H and MOUSE GNF1M for the corresponding species. The Present/Absent calls in the profiles were used to indicate whether a probeset was expressed or not, and the Marginal calls were also treated as Present calls. When a gene had many probesets, it was considered to be expressed if any one of the probeset was present. We ignored the probesets such as '_f_at', '_s_at' and '_x_at' because they could not be mapped to unique genes in a gene family. For a duplicated pair, common probesets shared by the two members were also excluded. Finally, if s was the number of tissues where both members were expressed, and d was the number of tissues where one member was expressed while the other was not, then their expression similarity was calculated as s/(s+d) [see additional file 1, 2, 3, and 4].
We identified 1,210 tandem duplicates and 410 retrogenes in the human genome, and 1,802 tandem duplicates and 680 retrogenes in the mouse genome. The higher number of tandem duplicates than retrogenes in both genomes implies a higher gene duplicability for tandem duplicates. Previous studies have found that gene duplicability is positively correlated with gene dosage  and gene complexity , although the correlation with functional essentiality is not always the same in yeasts and mammals [22-25]. To investigate the difference in gene duplicability between tandem duplicates and retrogenes in more detail, we counted the number of each type of duplicates in gene families with various sizes (Figure (Figure1).1). The result shows that their distributions among gene families are quite different (p < 0.01 for both genomes, chi-square test). Specifically, tandem duplicates are more likely to be enriched in larger families, whereas retrogenes do not display a preference.
Similarity of coding sequences has been widely used to indicate whether the new duplicates undergo gene conservation or functional divergence. While some reports have suggested that the duplicates really undergo sequence divergence when they are newly produced [9,26], other reports have found that they still remain more conserved than singletons . However, taking all duplicates as a whole will neglect some specific factors that belong to different molecular bases. For example, the effect of gene conversion, which keeps duplicates appearing similar through local DNA recombination , may have greater influence on tandem duplicates than retrogenes. The higher duplicability of tandem duplicates may also leave more recent and less divergent gene pairs. To test the hypothesis, we compared the amino acid identity between both types of duplicates (Figure (Figure2).2). The result shows that, the sequence identity of tandem duplicates is significantly higher than that of retrogenes (human: p = 0.021, mouse: p = 0.034, rank sum test). In agreement with Jun et al. , this result implies that tandem duplicates tend to be more conserved, whereas retrogenes tend to be more divergent.
In addition to the coding sequences, the evolution of regulatory elements is also important to determine the fate of duplicates. In fact, the differentiation of regulatory motifs can increase the expression specificity of the duplicates among various tissues and developmental stages, which is perhaps the most common form of subfunctionalization . Previous reports have found that a rapid expression divergence exists between duplicates , and that the expression diversity is also increased compared to singletons . However, as tandem duplications directly occur at the DNA level, it is more likely that the new duplicates preserve their original regulatory motifs and expression patterns. In contrast, as retrogenes are randomly inserted into the genome via mRNAs, they are more likely to acquire distinct regulatory motifs and expression patterns. To test this hypothesis, we compared the expression similarity for tandem duplicates and retrogenes by using microarray data across diverse tissues (Figure (Figure3).3). Although a lot of duplicates have been quite differentiated for both types, the expression similarity between tandem duplicates is still significantly higher than that between retrogenes (human: p = 0.036, mouse: p = 0.002, rank sum test). Therefore, the gene expression profiles also support the difference in functional divergence for both types of duplications.
As mentioned in the section Background, the retention mechanisms are determined by both functional divergence and evolutionary forces. To compare the evolutionary forces for both types of duplicates, we first performed the traditional dN/dS analysis (Figure (Figure4A).4A). The result shows little difference in the dN/dS ratios between tandem duplicates and retrogenes (human: p = 0.607, mouse: p = 0.257, t-test). In addition, the non-synonymous substitutions in most duplicates are under selective constraints (dN/dS < 1).
Nonetheless, the dN/dS test is directed to single site substitutions, which is not suitable for the case of whole gene substitutions such as the addition of duplicates. Lynch  has presented a new strategy for this issue by examining the role of effective population size. Briefly, if the new duplicates are nearly neutral and fixed by genetic drift, a small population size is favourable for their retention. On the contrary, if the new duplicates are advantageous and fixed by positive selection, the opposite should be true. In fact, Lynch has suggested that the long-term increase of duplicates from prokaryotes to eukaryotes is initially a neutral process in response to the reduction of population size . However, Shiu et al.  have argued that positive selection also plays an important role at least in mammalian genomes because there are more duplicates retained in the mouse lineage (larger population size) than in the human lineage (smaller population size), which cannot be explained by the difference in their duplication rate. Furthermore, since the generation time in mice is shorter than in humans, there will be more generations that are subject to selective pressures for mice and consequently, more duplicates retained in the mouse genome. In our dataset, there are both more tandem duplicates and more retrogenes in the mouse genome. To test if the excessive duplicates are really created in the mouse lineage, we grouped the age of the duplicates (inferred from dS) according to the divergence time between the two species (Figure (Figure4B).4B). The result shows that, while the duplicates generated prior to the split of the two genomes are more or less the same, there are more duplicates arising in the mouse-specific lineage (p < 1e-4 for both types, chi-square test). Based on the same assumption with Shiu et al. , this result implies that positive selection plays essential roles in the retention of both types of duplicates.
Of the two key dimensions to determine the retention mechanisms, we have found that the extent of functional divergence is distinct for tandem duplicates and retrogenes, whereas the underlying evolutionary forces are the same. As tandem duplicates are generated at the DNA level and easily influenced by gene conversion, they are more likely to be maintained (Figure (Figure22 and and3).3). Two main models can be used to account for the conservation of duplicates, i.e. dosage model and buffering model. The former proposes that as the new duplicates will increase the gene dosage, they can bring about some selective advantages . In contrast, the latter argues that the conserved duplicates are just used for compensation in case of the functional loss of their counterparts , and thus they are free from selective pressures. Given the signature of positive selection (Figure (Figure4),4), we propose that the dosage model is more prevalent in the fixation of tandem duplicates. In fact, the dosage model predicts that the fitness of dosage-sensitive genes will increase with the increase of gene copies , which is consistent with our observation that tandem duplicates tend to form large families (Figure (Figure1).1). Another large-scale functional analysis has revealed that tandem duplicates are enriched in receptors and binding proteins , which are also dosage-sensitive genes . Interestingly, copy number variants (CNV), which are strongly associated with segmental tandem duplicates , may also be maintained by dosage effect and positive selection .
Retrogenes and tandem duplicates display nearly opposite molecular properties. Since retrogenes are often distant from their parental counterparts and lose the original regulatory elements, they are more likely to undergo functional divergence (Figure (Figure22 and and3).3). There are also two main models available to account for the functional divergence, namely 'escape from adaptive conflict' (EAC) model  and 'duplication-degeneration-complementation' (DDC) model . Both of the models predict that the new duplicates will share the functions of the ancestral genes. However, the EAC model argues that duplications can release the potential benefits through functional specialization, whereas the DDC model only requires that the joint effect of the duplicates fulfil the original functions. The signature of positive selection in the retention of retrogenes votes for the prevalence of the EAC model (Figure (Figure4).4). In addition to our results, the analysis of gene movements has revealed that the X-linked genes are excessively transferred to autosomes via retropositions in mammalian genomes . These retrogenes can not only sustain essential functions during the inactivation of the male X chromosome, but also develop male-specific expression patterns [16,39]. The coexistence of functional divergence and selective benefits provides an important evidence for the EAC model.
The authors declare that they have no competing interests.
ZW conceived and performed the experiments. XD participated in the discussions. ZW and GD collected the data. ZW, GD and YL wrote and revised the manuscript.
All authors read and approved the final manuscript.
Human tandem duplicates.
Mouse tandem duplicates.
This research was supported by grants from National High-Tech R&D Program (863) (2006AA02Z334, 2007DFA31040), State key basic research program (973) (2006CB910705, 2010CB529206), Research Program of CAS (KSCX2-YW-R-112, KSCX2-YW-R-190), National Natural Science Foundation of China (30900272) and Special Start-up Fund for CAS President Award Winner (to G. Ding).